Composite quantitative and qualitative forecasting of futures prices: using one commodity to help forecast another

The thesis is composed of two chapters. The first chapter examines whether commodity price forecasting model performance can be improved by the inclusion of price forecasts for other commodities within the model specification. Using Bayesian Model Averaging methodology, we estimate 1330 different models to forecast the prices of hog, cattle, corn, and soybean and find strong support for the inclusion of one or more other commodity price forecasts in the best forecasting models.Also, sometimes the most important forecasting component is simply whether the price will move up or down. Such binary forecasts are commonly referred to as qualitative forecasts. The second chapter investigates whether qualitative forecasting of commodity prices can be improved by the inclusion within the model specification of price forecasts for other commodities. We estimate 1330 different models to forecast the price movements of hog, cattle, corn, and soybean and find strong support for the inclusion of one or more other commodity price forecasts in the best forecasting models as well. The results for both quantitative and qualitative forecasting suggest more work is called for to determine how best to use other commodity price forecasts to improve forecasting performance.